Very interesting, thanks all. (01)
A whole bunch of primary questions emerging here -- very "ontological', if I
may say... (02)
So for me, a couple of primary guiding themes: (03)
1) Is there any "one best way" to view this issue? Is there a simplest or most
parsimonious parsing of this process, in some simplest or minimalist feature
set, that enables us to construct this
classification/identity/similarity/difference framework? Yes, there are
clearly "many" ways to do this -- and those ways are not necessarily somehow
"wrong". But maybe (?) they are inherently limited or limiting or "localized"
in some way -- as suggested by their relative incommensurateness -- because
their internal mathematics lacks a simple fluency that enables in a simplest
and most natural way a full expression and development of all the facets of
this process -- of which there are clearly many. So, this issue of "simplest"
becomes a definition for "optimizing" -- and this talk of "mandating" reflects
an instinct for optimal simplicity and clarification. Yes, there are many
ways. But is there one best way (as defined by various criteria of "best")? (04)
2) What is the most natural and authentic and confirmed model of general
cognitive processing -- is there a best way to understand what human beings are
actually doing? Of course, the psychological literature is rich with various
approaches. But maybe there is a generalization of this range of ideas that
can embrace them all (as per the famous Smith and Medin on "Categories and
Concepts") such that we don't have to struggle with some either/or choice among
psychological models of classification -- but can simply interpret all these
approaches as alternative facets of a single containing framework. See
http://originresearch.com/docs/SmithAndMedin.docx for the first three chapters. (05)
3) The fundamental tension in this discussion -- I would say -- is the split
once known as "the holy war between the scruffies and the neats" --
https://en.wikipedia.org/wiki/Neats_vs._scruffies -- which I tend to
generalize simply as the tension between bottom-up empirical-driven "local"
approaches, and top-down elegance-and-simplicity-driven "global" models, which
attempt generalization of the empirical frameworks as special cases -- and
which, to date, as JS will remind us, have not succeeded. To date, global
models do not map smoothly to and "contain all" local models. Is that fact a
property of the universe and "the logos" or somehow inherent in any possible
language -- or an artifact of a mathematical science that is not quite mature?
The Wikipedia article says "Much success in AI came from combining neat and
scruffy approaches." I think that's the right way to see the issue today, and
a direction worth pursuing. (06)
REAL OBJECT - ABSTRACT OBJECT (07)
My instinct is to insist on a couple of methodological preliminaries -- having
to do with the nature of "objects" -- meaning "real objects" (which, as JS and
many others have noted, have blurry and ill-defined or non-existent boundaries
that can only be defined by context-specific motivation) -- and the "abstract
objects" that represent those real objects in some computing medium. (08)
Way back in the day, when I was first driving into this territory, and looking
for ways to build a comprehensive and algebraically-integrated model of
epistemology based on a general theory of "concepts", I came across the book
Programming Languages, Information Structures, and Machine Organization, by
Peter Wegner (1968), then a textbook in computer science at UC San Diego, and,
I believe, his PhD thesis. After banging through that rather amazing and
comprehensive book as best I could, I became convinced that the most reasonable
way to understand "concepts" is as "information structures". That's the right
way to understand "what a concept is" -- and the right way to construct an
"abstract object". So -- everything we are going to do to build a model of the
world is going to be optimally defined, in the least confusing way, with the
fewest number of intervening levels and layers -- by defining absolutely every
element of our definition system as "information structures" -- beginning with
bits and bytes, and expanding to things like rows and columns ("row vectors")
-- and always "linearly optimizing" the mapping of this abstraction hierarchy
to the physical ground of the "machine space" -- the actual electronics. "Zero
distance" as the optimizing variable. "Make a perfect linear map." Abstract
concepts like "similarity" or "difference" or "identity" -- or "analogy" or
"comparison" -- are going to be defined in these terms. Always, we are
discussing the properties of abstract models. (09)
I had been looking at the entire issue of "what is a modeling language?" -- and
how can these "bony structured concepts" (these abstract linear objects called
"concepts") form a smoothly isomorphic map of a continuous reality (for
example: fluid dynamics and the flow of liquids)? And what does the "hierarchy
of computer languages" have to do with this process, if anything (ie,
high-level languages are defined in terms of lower-level languages, with
intervening "macros" that assemble the hierarchy of high-level functions)? The
way it looks to me, bottom-level machine code is the "atomic" level of the
coding hierarchy, and the correct or optimal ground for any attempt to build an
accurate model of continuous variation in a digital medium. So, to approach
continuous variation in the model, in its machine representation maintain a
smooth code mapping directly to machine code -- to the "zeroes and ones". (010)
For perfection and minimization, I want to remove any "knots" or weak/confusing
or ambiguous "homomorphic" maps in the intervening levels of this language
hierarchy. I don't want any definitions that are not rigidly/perfectly mapped
to the lower level language, if possible -- so that the entire cascade of
abstract representations, from machine code to the highest level composite
macros, is smoothly and directly mapped. No bumps along the way and no
interpretive ambiguity or uncertainty. Perfect that process, and pull out all
the weeds. We want high-level composite abstractions -- with absolutely smooth
and "near-continuous" maps directly to the ground of finite-state on/off
machine code. (011)
THE ABSOLUTE PRIMACY OF MACHINE REPRESENATION (012)
This approach gives an absolute primacy in the modeling process to machine
representation. It says -- yes, we start with observation -- but our target is
perfect machine representation -- so don't compromise that method, or your
system will get "dirty" -- and if it does, your logic flow will get jammed up
-- and you will be forced to build a world of ad-hoc special-case local systems
that can't talk to each other.... (013)
> It's irrelevant how you represent the properties or what conventions
> you adopt for storing information about them.
>
> You still have to observe the patterns before you can *infer* whether
> or not they determine a unique item. (014)
Yes -- but your "observed patterns" ARE a "model" -- constructed in terms of a
presumed underlying language -- so even your "objective perception" is an
"inference". So -- I would take an opposite view -- and say it is essential to
discipline the representations of the perceived properties (carefully choose
the way you define those properties) and your conventions for storing
information. If you introduce muddiness or ambiguity at this point in the
modeling/representation process, everything that follows risks muddiness... (015)
Bruce Schuman, Santa Barbara
http://networknation.net/global/vision.cfm (016)
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Alexander Titov
Sent: Wednesday, June 17, 2015 3:42 AM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] The "qua-entities" paradigm (017)
An identity - as a specific classification of a thing in a such way that the
correspondent class has one and only one member? (018)
And I think that an identity seriously depends on a viewpoint and an observer
who/which makes that identity ‘classification’. (019)
Regards,
Alex
> On 17 Jun 2015, at 10:25, Matthew West <dr.matthew.west@xxxxxxxxx> wrote:
>
> Dear Rich,
> Each grain of sand exists in the real world and has identity, whether
> or not you are interested in them. That is something entirely
> different. A handful of sand is also something that exists in the real
> world (the aggregate of the grains of sand whilst they are in your
> hand) and whether you care about that is also a different question.
>
> Regards
>
> Matthew West
> Information Junction
> Mobile: +44 750 3385279
> Skype: dr.matthew.west
> matthew.west@xxxxxxxxxxxxxxxxxxxxxxxxx
> http://www.informationjunction.co.uk/
> https://www.matthew-west.org.uk/
> This email originates from Information Junction Ltd. Registered in
> England and Wales No. 6632177.
> Registered office: 8 Ennismore Close, Letchworth Garden City,
> Hertfordshire,
> SG6 2SU.
>
>
>
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
> [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich
> Cooper
> Sent: 17 June 2015 06:49
> To: '[ontolog-forum] '
> Subject: Re: [ontolog-forum] The "qua-entities" paradigm
>
> Are you saying that identity must *always* be *unique*? I can
> identify a handful of sand at the beach without assigning an identity to each
>grain.
> All grains look the same to me, therefore all sand has the same
> identity, so I treat it as a unitless object, and the best I can do to
> subdivide it is to organize it into specific volumes, weights and prices.
>
> Sincerely,
> Rich Cooper,
>
> Chief Technology Officer,
> MetaSemantics Corporation
> MetaSemantics AT EnglishLogicKernel DOT com ( 9 4 9 ) 5 2 5-5 7 1 2
> http://www.EnglishLogicKernel.com
>
>
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
> [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F
> Sowa
> Sent: Tuesday, June 16, 2015 10:30 PM
> To: ontolog-forum@xxxxxxxxxxxxxxxx
> Subject: Re: [ontolog-forum] The "qua-entities" paradigm
>
> On 6/17/2015 1:12 AM, Rich Cooper wrote:
>> you could say that the ID is the concatenated value of all
> properties
>
> I was trying to explain that similarity is observable, but identity is
> always an inference.
>
> It's irrelevant how you represent the properties or what conventions
> you adopt for storing information about them.
>
> You still have to observe the patterns before you can *infer* whether
> or not they determine a unique item.
>
> John
>
> (020)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (021)
|